An Efficient Implementation of File Sharing Systems on the ... - CiteSeerX

5 downloads 72 Views 336KB Size Report
DHT based Chord [7] [8], Pastry [9], Tapestry [10] & content based CAN [11]. ... networking theories and technologies, Liu suggested that a new computing ...
An Efficient Implementation of File Sharing∗ Systems on the Basis of WiMAX and Wi-Fi Jingyuan Li, Liusheng Huang, Weijia Jia, Mingjun Xiao and Peng Du Joint Advanced Research Institute of City University of Hong Kong and University of Sci. Tech. China Suzhou, China

Abstract—This paper proposes an efficient algorithm for P2P file sharing systems based on WiMAX mesh mode and Wi-Fi technologies. Wireless networks in our system are hierarchically divided into three layers: Wi-Fi based wireless local area networks under a subscriber station; the mesh network of subscriber stations under a base station; the network of base stations. File lookup procedure may go through three steps: the requesting end host firstly looks up for the requested file within it’s neighbor end hosts under the same subscribe station; if fails, then it sends lookup messages to the nearest subscriber stations in the mesh network following to a changed dynamic source routing protocol; if both the steps fail, then the requesting end host searches the requested file through the Chord-based base stations’ overlay. By statistical analysis and simulations, we prove that our layered P2P file sharing system can reduce the number of lookup messages in physical networks and prevent over expenses of precious bandwidth in wireless metropolitan area networks. Index Terms—P2P File sharing system, WiMAX

I. INTRODUCTION P2P file sharing systems are growing extraordinarily fast in the past decade: 30 percent of the internet traffic is peer-to-peer related by the end of the year 2004 [1]. There are many successful applications in P2P domain including parallel computing [2], collaborative learning [3], resource sharing etc. Among these application domains, P2P file sharing is the most popular application despite of the lawsuits about piracy to music and movies sharing providers [4]. P2P file sharing systems should at least implement three protocols: file-release protocol, file-lookup * The research described here is partially supported by CityU Research Project - Infrastructure Account of Faculty/School/College (APR) no. 9610027 and the National Basic Research Fund of China (“973” Program) under Grant No.2003CB317003.

1-4244-0507-6/06/$20.00 ©2006 IEEE

protocol and file-transportation protocol. In this paper, ‘file’ and ‘resource’ are interchangeable terms, i.e., ‘file lookup’ and ‘resource lookup’ are treated as the same. Based on the implementation of file-lookup protocol, P2P file sharing systems are classified into two categories [5]: unstructured lookup protocols such as flooding [6] and structured lookup protocols such as DHT based Chord [7] [8], Pastry [9], Tapestry [10] & content based CAN [11]. Meanwhile, wireless techniques are developing in an amazingly fast speed. Wireless networks including both circuit networks and packet networks are ubiquitous throughout the world. In fact, the goal of accessing network anytime anywhere will be fully achieved in the near future. Among the wireless standards, WiMAX [12] [13] [14] is a new and exciting member: it designs the wireless metropolitan area networks (WMAN) successfully. We are going to describe WiMAX standard in detail in section III. Both P2P file sharing systems and the WMAN standard WiMAX have a brilliant future. Therefore, combining the two techniques surely will bring the benefits to the users of both sides. However, the problem is: “do the facilities of WiMAX efficiently support the structures of file sharing systems?”. Unfortunately, the answer is not promisingly positive. The conflict of inefficiency in pure P2P file sharing systems (P2P systems that completely omit physical network topologies), lack of air interfaces and energy resources are the most severe problems that fundamentally retard the growth of P2P applications, especially P2P file sharing systems on wireless networks. Some works have been done to solve these problems [15] [16] [17]. However, no researcher ever presented any solution for efficiently implementing the most popular DHT based system Chord on WiMAX motivated wireless metropolitan area networks. In this paper, we propose a layered P2P file sharing structure on WiMAX and Wi-Fi network infrastructures. By simulations, we prove that

819

our hybrid structure is easy to construct and efficient to maintain with a little extra cost. The rest of the paper is organized as follows: section II presents some related work; section III gives a fundamental introduction to WiMAX and its mesh mode; section IV proposes our hybrid P2P file sharing system with support of WiMAX and Wi-Fi. We analyze and simulate our system in section V and section VI respectively. Section VII concludes the paper.

control method that efficiently protected wireless enddevices from hard workload. Meanwhile, Andersen [23] et al. proposed an architectural concept for mobile P2P file sharing services that narrowed the gap between hierarchically structured cellular communication networks and non-hierarchically structured P2P applications. III. WIMAX OVERVIEW AND MESH MODE

II. RELATED WORK In recent works, much attention has been focused on peer-to-peer applications. A review by Milojicic and Kalogeraki [5] suggested that P2P applications be divided into three dimensions: file sharing, distributed computing and communication-&collaboration. Nonetheless, distributed computing and communication&- collaboration have not been developed well due to some technical problems which are hard to be conquered. On the other hand, P2P file sharing systems are very successful. Many P2P file sharing protocols like CAN [11], Chord [7], Pastry [9], eDonkey [18] etc. are well implemented and are being used by a large number of people all over the world. Among the P2P file sharing systems, Peng [15] suggested that the system has two major functions: resource lookup and resource transportation. For resource lookup function, Stoica et al. proposed a scalable P2P lookup protocol named Chord [7] for internet applications. They introduced the distributed hash table [19] technology to their system to provide the lookup operation: given a key, Chord system maps the key to a node. Data location is easily implemented in Chord by associating a key with each data item and storing the key-data pair at the mapped node. However, Chord does not support keyword lookup. Recognizing this limitation, Zhu and Hu [20] established an architecture for content-based publish/subscribe services that perfectly solved the problem. For resource transportation function, Ghandeharizadeh [21] et al. investigated a novel streaming architecture consisting of home-to-home online devices that collaborated with one another to provide on-demand access to large repositories of continuous media such as audio and video clips. At the same time, with the development of wireless networking theories and technologies, Liu suggested that a new computing paradigm called mobile computing that has been emerged as an important research field in which users are able to access resources and services with mobile devices. Nevertheless, some characteristics of wireless computing make wireless P2P hard to implement. Leung and Kwok [22] designed a topology

A. WiMAX Overview WiMAX [12] forum was founded in April 2002 by leading vendors of wireless access equipment and telecommunications components. The forum’s mission is to lay the groundwork for an industry-wide acceptance and implementation of the IEEE 802.16 and ETSI HiperMAN standards likewise the Wi-Fi association and DSL forum did in wireless local area network and broadband wire-line domain respectively. WiMAX stands for Worldwide Interoperability for Microwave Access. It is a standard for wireless metropolitan area network which provides the compatibility of IEEE 802.16 and HiperMAN. WiMAX provides a mechanism for last mile wireless access of the network filling in the gap between wired network access facilities, as depicted in Fig. 1. Subscribe Station Subscribe Station

Internet/PSTN Subscribe Station Subscribe Station

Base Station

Subscribe Station

Subscribe Station

Fig. 1 WiMAX structures

WiMAX MAC layer standard provides QoS assured scheduling mechanisms in base stations (BS) and subscriber stations (SS), with different applications being provided with different qualities of services. In MAC layer standard, there are two different network architectures: the point-to-multipoint (PMP) mode and the mesh mode. The main difference between the PMP and optional mesh mode is that, in the PMP mode, traffic only occurs between the BS and SSs while in the mesh mode, traffic can be routed through other SSs and can occur directly between SSs. SSs mesh mode scheduling algorithms are divided into three categories: distributed

820

scheduling algorithms, centralized scheduling algorithms and a combination of previous two algorithms.

sections. Overall structure of our layered P2P file sharing system is depicted in Fig. 3.

B. WiMAX Mesh Mode WiMAX mesh mode [24] is fairly a new topic provided by WiMAX forum. The structure of the mesh mode is depicted in Fig. 2.

BS2

DHT based BS network SS mesh network

BS

BS1

BS3

SS mesh network

BS

Backhaul Networks End host s

End host d

Fig. 3 Overview of our layered file sharing system BS

BS

Fig. 2 Mesh-based WiMAX infrastructure

The architecture is an ad-hoc like structure in which SSs act as the routers. Traffic is routed through other SSs or occurs directly among the SSs. It should be noted that, there is no need for the SSs to connect directly to the BS. We suppose that Wi-Fi technology is used to connect end hosts to the corresponding SS and the backhaul of the network is wired.

B. File Publication When an end host s releases a file f to the file sharing system, it registers its information in two places: the SS the end host is directed connected to and a BS in the BS network. This is shown in Fig. 4. BS2 h(f)=h(BS2) DHT based BS network SS mesh network

BS1

BS3

SS mesh network

IV. OUR PROPOSED FILE SHARING SYSTEM A. Overview of Our File Sharing System We come to a major problem: how can we efficiently design the P2P file sharing systems based on the wireless network infrastructures mentioned above? To solve this problem, we propose a hybrid P2P file sharing system which is aware of the underground network topologies. End devices under the same SS are connected with Wi-Fi technology, i.e., every two end hosts are one-hop connected and each end host in the same local area network is connected directly to the SS. At the same time, we propose a changed dynamic source routing (CDSR) lookup protocol on SSs mesh networks where the number of SSs under a BS is not large enough to efficiently construct a DHT based file sharing architecture. Moreover, by implementing CDSR on SSs mesh networks, our layered file sharing system can support keyword lookup within a SSs mesh network. We also design DHT based file lookup architecture on BSs networks, because BSs networks are cable connected and able to tolerate the extra traffic loads. We will describe our systems in more detail in the following

f, keywords

End host d

End host s

Fig. 4. File publication in our layered file sharing system.

Firstly, s releases the file name and a bunch of keywords to the SS to which it is registered. By doing so, we guarantee that our system support content lookup within the same range of a BS. Secondly, s computes the hash value of the file name h(f) using some universal distributed hash functions such as SHA-1 and registers the file to the corresponding BS in the BS network, where h(f)=h(BSi) and BSi is the ID of the corresponding BS. If such BSi does not exist, then the information is stored in the nearest next BS as that is implemented in the Chord [7] system. Registering f to the DHT based BS overlay network enables the end hosts from other BSs to share their resources together in a robust and efficient way.

821

C. File Lookup When an end host d requests a file f, it may require three performance steps to finally find the requested resource as depicted in Fig. 5. BS2

If there is no f found in the local SSs mesh network, then the resource lookup procedure sends a request message to the DHT ring of the BSs network as Chord does, seeking fortune from hosts under other BSs. The whole lookup procedure is described in the following flow chart in Fig. 7. Start

DHT based BS network SS mesh network

BS1

BS3

( Directly send message to neighbors )

SS mesh network

Y

Whether f is under the same SS

N

Search f in the BS network

Search f in the local SS mesh network based on DSR

End host d

Y

Whether f is under the SSs mesh network

(CDSR)

N (Chord based BSs ring)

Search f in BSs network

End host s

Fig. 5. File lookup in our layered file sharing system

Finish

Firstly, d searches f on end hosts within the SS it is registered. Because all of the end devices are one-hop connected, d instantly knows whether f exits or not. If f does not exist within the local area network under the registering SS, then d searches the publication information within the range of the corresponding BS, i.e., the nearest SSs mesh network. Here a changed dynamic source routing protocol (CDSR) is used to efficiently manipulate the routing problems in the SSs network as shown in Fig. 6.

Fig. 7. Flow chart of the lookup procedure

By importing the above three layer lookup procedure, our layered P2P file sharing system guarantees the following two advantages. Firstly, the requesting end host gets the requested file as ‘near’ as possible. This saves a great amount of bandwidth for the underground wireless infrastructures. Secondly, our system supports keywords lookup within the SSs mesh networks. On the basis of our simulations, this may be the major position that the requested file f is found. D. File Transportation Resource transportation is almost the same as that of other cabled networks as depicted in Fig. 8. BS2

Fig. 6. Changed dynamic source routing (CDSR) DHT based BS network

CDSR is very similar to DSR algorithm [25] in mobile ad-hoc networks except that most of the SSs are immobile as compared to the end devices in the ad-hoc networks. The requesting end host (the destination end host) d launches a request procedure in the SS to which it is registered in order to find f. The SS sends a request message containing the name of the file f and/or some keywords to all its neighbors in the SSs mesh network. The neighbor SSs perform the same action until a SS finds the file f in its resource register table, or all the SSs under the range of the same BS receive the message and no match is found. If f is found, then the source SS unicasts backwards to the requesting end host where the requesting end host saves the routing information in its routing table.

SS mesh network

BS1

BS3

SS mesh network

Search f in the BS network

Search f in the local SS mesh network based on DSR

End host d

End host s

Fig. 8 File transportation in our layered file sharing system

V. ANALYSIS In this section, we statistically analyze the performance of our layered P2P file sharing system.

822

Suppose there are ¢ end devices under each SS, £ SSs under each BS and ¤ BSs of the whole system. Let ¨ be the probability of file f on each end host. In our hybrid P2P file sharing system, the probability pEH that a requesting end device obtains the requested file f under the same SS is shown in expression (1).

p EH = 1 − (1 − η ) α

(1) Defining pSS as the probability of the requesting end host obtaining the requested file f under the same BS or in the nearest SSs mesh network (but not in its own Wi-Fi based local area network). pSS is shown in expression (2).

p SS = (1 − η )α × [1 − (1 − η ) α ( β −1) ]

Fig. 9 Percentage of three types of requested file positions

(2) pBS is the probability indicating that the nearest SSs mesh network does not have the file transmission request is shown in expression (3).

The above figure clearly presents the fact that, when a file f is popular, a great number of end hosts cached it to thir own memory. Therefore the requesting end host is αβ more likely to obtain f around its neighbors in our p BS = 1 − (1 − η ) (3) hierarchical system which saves a great amount of From the above analysis, we come to the precious wireless bandwidth. We also notice that even conclusion: as the popularity of resource goes higher, though a file is very unpopular (for example, only 0.5% there is more opportunity that the requesting end host end hosts have the file), there is still a great probability obtains the requested file within the range of the SS to that the requesting end host gains the file within the SSs which it is registered and the nearest SSs mesh network. mesh network where it is located (45% in the above This property enhances the performance of our layered simulation result), as long as there are enough end hosts P2P file sharing system, reducing both the number of under the corresponding BS. lookup messages in file lookup procedure and the Fig. 10 presents the relationship between the number bandwidth requirement in WiMAX and Wi-Fi based of lookup messages in the physical networks and the wireless network architectures in file transportation popularity of the requested file in our hierarchical procedure. system. The number of lookup messages are greatly reduced as the file becomes more and more popular. In original Chord system, the popularity of the requested VI. SIMULATIONS file does not affect the complexity of the lookup We simulated our layered P2P file sharing system procedure. Our system successfully improves this under an independent WiMAX wireless metropolitan inefficiency with only a little extra architecture built on area network with 100 BSs connected by cables. Each WiMAX infrastructures. BS controlled 25 SSs, all of which were interconnected to form a wireless SSs mesh network. The network under each SS was set to be Wi-Fi based local area network. End hosts under the same SS were one-hop connected and were directly connected to the SS they were registered to. Four end hosts were registered to each SS. The popularity of a file f was from 0.5% to 6%. Fig. 9 describes the percentage of three types of positions of the requested file f: source end host which was under the same SS (Wi-Fi based local area network) with the destination end host; source end host which was under the same BS (SSs mesh network); source end host Fig. 10 The number of lookup messages in the physical network which was under other BS (BSs Chord ring). 













823









  ILOHSRSXODULW\ 











VII. CONCLUSION We have proposed a layered P2P file sharing system on the important WiMAX and Wi-Fi wireless techniques. Requesting end host obtains the requested file from three types of positions: it first searches neighbor end hosts under the same SS; if this step does not find the requested file, then the requesting end host sends lookup messages within the nearest SSs mesh network using CDSR protocol; if all the above steps fail, the requesting end host launches a lookup process on Chord based BSs ring overlay network. By statistical analysis and practical simulations, we have shown that our layered P2P file sharing system can greatly reduces the number of lookup messages in physical networks and prevents over expenses of precious bandwidth in wireless metropolitan area networks. Our future work aims at implementing the system with consideration of the mobility of wireless end hosts and also combining wireless P2P file sharing architecture along with more powerful and stable cable network P2P applications.

REFERENCES [1] Top Tech News, http://www.toptechnews.com/story.xhtml?story_id=38121, 2005. [2] G. Chen, “Parallel Computing: Algorithms and Practices”, High Education Press, 2003. [3] I. Vedula, R. Han, “A distributed software system architecture for wireless Peer-to-Peer collaborative learning”, The 3rd IEEE International Conference on Advanced Learning Technologies, 2003, pp. 7-11. [4] T. Karagiannis, M. Faloutsos, “P2P Technology Still Going Strong Despite Lawsuits, Fines”, http://www.newsroom.ucr.edu/cgi-bin/display.cgi?id=935, Dec 2004. [5] D. S. Milojicic et al. (2002). “Peer-to-Peer Computing”, Technical Report HPL-2002-57, HP Lab, 2002. [6] Gnutella, http://www.gnutella.com. [7] I.Stoica, R. Morris, D. Karger, F. Kaashoek, and H. Balakrishnan, “Chord: A peer-to-peer lookup service for Internet applications”, presented at the ACM SIGCOMM, San Diego, CA, Sep. 2001. [8] A. Singh, L. Liu, “A hybrid topology architecture for P2P systems”, Computer Communications and Networks, 2004, pp. 475-480. [9] A. Rowstron and P. Druschel, “Pastry: Scalable, distributed object location and routing for large-scale peer-to-peer systems” in International Conference on Distributed Systems Platforms, 2001. [10] B. Y. Zhao, L. Huang, and J. Stribling, “Tapestry: a resilient global-scale overlay for service deployment”, IEEE Journal on Selected Areas in Communications, Volume 22, Issue 1, Jan. 2004, pp. 41-53. [11] S. Ratnaswamy, P. Francis, M. Handley, R. Karp, and S. Shenker, “A scalable content-addressable network”, in Proc. IEEE INFOCOM, 2002, pp. 1190-1199. [12] 802.16-2004, IEEE Standard for local and metropolitan area networks, 2004.

[13] W. Du, W. Jia and G. Wang, “A neighborhood-based bandwidth scheduling scheme in WiMAX networks”, Journal of Software, Volume 17, No.2, February 2006, pp. 267-274. [14] WiMAX Forum, http://www.wimaxforum.org/home/. [15] G. Peng, S. Li, H. Jin and T. Ma, “M-CAN: a lookup protocol for mobile peer-to-peer environment”, Parallel Architectures, Algorithms and Networks, 2004, pp. 544-549. [16] H. Wolf, M. Wang, “A framework with a peer fostering mechanism for mobile P2P game development”, Mobile Business, 2005, pp. 109-115. [17] F. Araujo, L. Rodrigues and J. Kaiser, “CHR: a distributed hash table for wireless ad hoc networks”, 25th IEEE International Conference on Distributed Computing Systems Workshops, 2005, pp. 407-413. [18] eDonkey, http://www.edonkey2000.com/. [19] W.E. Burr, “Cryptographic hash standards: where do we go from here?” Security and Privacy Magazine, IEEE, 2006, pp.8891. [20] Y. Zhu, and Y. Hu, “Ferry: an architecture for content-based publish/subscribe services on P2P networks”, International Conference on Parallel Processing, 2005, pp. 427 – 434. [21] S. Ghandeharizadeh, B. Krishnamachari and S. Song, “Placement of continuous media in wireless peer-to-peer networks”, IEEE Transactions on Multimedia, Volume 6, Issue 2, April 2004, pp. 335 – 342. [22] A. K. Leung and Y. Kwok, “On Topology Control of Wireless Peer-to-Peer File Sharing Networks: Energy Efficiency, Fairness and Incentive”, World of Wireless Mobile and Multimedia Networks, 2005. Sixth IEEE International Symposium on a 1316 June 2005, pp. 318 – 323. [23] F. Andersen, H. Meer and C. Kappler, “An Architecture Concept for Mobile P2P File Sharing Services”, Informatik 2004 – Informatik verbindet, pp. 229-233. [24] D.Beyer, N. Waes and C. Eklund, “Tutorial: 802.16 MAC Layer Mesh Extensions Overview”, NOKIA, 2002. [25] A. Boukerche, “Performance evaluation of routing protocols for ad hoc wireless networks”, Mobile Networks and Applications, Volume 9, Issue 4, 2004.

824